Runs Canek integration.
RunCanek(x, ...)
# S3 method for class 'Seurat'
RunCanek(
x,
batches = NULL,
slot = "data",
assay = NULL,
features = NULL,
selection.method = "vst",
nfeatures = 2000,
fvf.nfeatures = 2000,
integration.name = "Canek",
debug = FALSE,
correctEmbeddings = TRUE,
pcaDim = NULL,
maxLoop = 5,
loopTol = 0.001,
...
)
# S3 method for class 'SingleCellExperiment'
RunCanek(
x,
batches = NULL,
assay = "logcounts",
integration.name = "Canek",
debug = FALSE,
...
)
# S3 method for class 'list'
RunCanek(x, ...)object with expression counts or list of matrices.
additional arguments passed down to methods, e.g. ncores to parallelize
MNN pair finding (see CorrectBatches; defaults to 1, sequential).
for S4 objects the column containing batch information.
slot used for Seurat objects (default: data).
assay used for Seurat objects. If NULL (default), an existing "SCT" assay is preferred over the object's default assay.
optional vector of features to use for correction.
method used for FindVariableFeatures on Seurat objects when features is NULL.
number of features returned by SelectIntegrationFeatures.
number of features returned by FindVariableFeatures.
name for the integrated assay.
whether to store information about correction vector.
whether to perform the correction on PCA embeddings instead of gene expression (Seurat objects only).
number of PCA dimensions to use when correctEmbeddings is TRUE. If NULL (default), it is inferred from the object's existing "pca" reduction; if none is found, used 30 as default with a warning. When correctEmbeddings is TRUE and a "pca" reduction already exists on the object, it is reused directly instead of being recomputed. If the assay in use is SCTransform-normalized and no "pca" reduction exists, an error is raised instead of computing one internally, since SCTransform fit separately per batch produces residuals that are not on a comparable scale across batches.
number of times to iterate the correction (correctEmbeddings = TRUE only), using each iteration's corrected result as the input to the next. Defaults to 5.
average change in median correction magnitude used to stop iterating early. Defaults to 1e-3. Ignored if maxLoop = 1.
An object of the appropriate type.